{ "schema": "projectforty2.kram.terrain.v1", "public_skin_version": "kram-resident-sota-2026-07-17", "generated_at": "2026-07-17T00:00:00Z", "mission": "Kram sorts Alzheimer's and dementia research. It shows where the field is stuck and which tests could change its mind.", "membrane": [ "No diagnosis", "No treatment guidance", "No prevention/cure promises", "No private family data", "No fake citations", "Personal urgency may choose the question; evidence decides the answer" ], "response_renderer": [ "boundary", "terrain", "evidence", "uncertainty", "falsifier", "next_action" ], "resident_state": { "state": "resident", "label": "active resident research organ", "cadence": "continuous public intake, 6-hour hypothesis probes, monthly literature/audit resweeps", "role": "show what is known, what is overclaimed, what would change the map, and where research effort is missing" }, "sota_summary": { "headline": "Kram reads the field as a few strong ideas, each real but incomplete, not one solved story of the cause.", "plain_read": "Amyloid, tau, inflammation, vascular injury, metabolic signaling, infection, sleep, and endolysosomal routing all matter somewhere. Kram's job is to stop these signals from collapsing into one favorite story: each claim must name the cohort, assay, confound, falsifier, and next decisive test.", "current_working_thesis": "July 2026 update: the set of papers that carry a role grew 47-fold, from 94 to 4,425, and one of Kram's own earlier claims did not survive. Kram had said the field already holds more evidence for the other lanes than for amyloid. That claim is now shown false. About 2,126 role tags point to amyloid. About 1,581 point to the vascular, tau-seed and endolysosomal lanes together, even though the search was built to over-sample those lanes about 2.8 times. The wider read still holds: amyloid is real but not the whole cause (Kram's class R5b), and its escape-pressure, how hard the evidence pushes toward other causes, is 0.43. Kram's own tool corrected Kram. That is the point." }, "field_map": [ { "lane": "Amyloid / plaque burden", "state": "costumed scalar / amplifier", "quick_read": "Real biology and real drug effect, but current evidence fits an amplifier with a slowing ceiling rather than a complete root-cause explanation. About 2,126 role tags in the 4,425-paper role-carrying corpus point to amyloid, so it is also the single largest lane - dominance the earlier map understated.", "tracking": "Ceiling breaches, tau-PET slope effects, subgroup responses, downstream rescue without amyloid change.", "gap": "Separate symptomatic slowing, biomarker movement, and disease-root claims in the same analysis." }, { "lane": "Tau seed competence", "state": "independent axis candidate", "quick_read": "Bulk tau can behave like a severity marker; seed competence and propagation are sharper causal questions.", "tracking": "MAPT / GENFI / ALLFTD longitudinal panels, tau-PET propagation velocity, seed assays tied to upstream stressors.", "gap": "The named cohorts (ALLFTD, GENFI) do not collect the ERF marker panel the decisive test needs - the measurement, not the cohort, is missing." }, { "lane": "Cerebrovascular injury", "state": "escape-direction candidate", "quick_read": "White-matter hyperintensity, perfusion, BBB, and vascular-risk data create real pressure against single-axis amyloid stories - but the corpus no longer shows this lane out-massing amyloid, as an earlier read claimed.", "tracking": "WMH-stratified CLARITY-AD / TRAILBLAZER tau-PET slope analyses and microvascular confound controls.", "gap": "The most-named trial for this test (CLARITY-AD) records no tau endpoint and screens out severe white-matter disease - no registered study currently collects tau-PET slope, amyloid status, and WMH tertiles in one powered cohort." }, { "lane": "Endolysosomal / retromer routing", "state": "candidate convergence node", "quick_read": "SORL1, retromer, CatD, and endosomal traffic are where amyloid, synapse, lipid, and inflammatory stories can meet mechanistically.", "tracking": "DIAN pro-CatD temporal precedence, SORL1-retromer rescue, SEC fractionation, MDH2 anchor behavior.", "gap": "The decisive banked-CSF temporal-precedence analysis is gated behind a pro-cathepsin-D CSF assay that has zero published Alzheimer's precedent - the assay, not DIAN access, is the bottleneck." }, { "lane": "Neuroinflammation / complement / TREM2", "state": "mixed scalar and axis", "quick_read": "Inflammation often tracks damage, but microglial genetics and complement-synapse mechanisms can be causal in narrower contexts.", "tracking": "Synapse-preservation endpoints, complement inhibition, TREM2 stratification, cell-state timing.", "gap": "The complement test has no registered vehicle: the antibody named for it (ANX005) has no Alzheimer's program, and no complement-directed AD trial pairs a synaptic-PET endpoint with amyloid imaging." }, { "lane": "Metabolic / GLP-1 / insulin signaling", "state": "translational axis under test", "quick_read": "Metabolic interventions are plausible but often confounded by vascular risk, weight, glycemia, and survival effects.", "tracking": "Non-diabetic normoglycemic AD readouts, stratified GLP-1 trials, cognition vs biomarker divergence.", "gap": "The GLP-1 readout (EVOKE) already published in 2026 without a pre-specified normoglycemic subgroup and without collecting HbA1c - the trial cannot answer the mechanism question as posed." }, { "lane": "Infectious / immune-history hypotheses", "state": "marginal open question", "quick_read": "Signals such as zoster vaccination are interesting enough to keep alive, not strong enough to promote as causal closure.", "tracking": "Prospective mechanistic cohorts, pathogen-reactivation markers, quasi-experiment replication, negative controls.", "gap": "Move from epidemiologic association to mechanism-timed evidence." }, { "lane": "Sleep / glymphatic / production-clearance balance", "state": "contextual modulator", "quick_read": "Sleep links to amyloid and cognition, but directionality and production-versus-clearance dominance stay unsettled.", "tracking": "SILK kinetics, sleep intervention timing, production/clearance separation, vascular interaction.", "gap": "The one relevant human SILK dataset (Lucey 2018) computed production by assuming clearance is unchanged, so the production-vs-clearance ratio is undefined - the prediction as written may not be scorable from any single study." } ], "gap_board": [ { "id": "GAP-001", "priority": "decisive", "question": "Does pro-CatD/total inflect at EYO <= -25 in DIAN before Ab42/40 decline?", "why_it_matters": "Would promote or demote endolysosomal/retromer failure as an early convergence node.", "what_would_move_it": "A validated isoform-resolved pro-cathepsin-D CSF assay first (none exists in Alzheimer's), then banked DIAN CSF, SEC fractionation, matched mutation strata, MDH2 anchor." }, { "id": "GAP-002", "priority": "testable now", "question": "Does high baseline WMH carry more of the anti-amyloid tau-PET slope benefit?", "why_it_matters": "Tests whether vascular injury is a hidden response-stratifier rather than a nuisance covariate.", "what_would_move_it": "A trial that actually holds tau-PET slope AND WMH tertiles together - CLARITY-AD has neither tau nor high-WMH patients; TRAILBLAZER-ALZ 2 has tau-PET but no WMH stratifier. The vehicle has to be built or an ADNI-class cohort re-analyzed." }, { "id": "GAP-003", "priority": "watch", "question": "Can complement inhibition preserve synapses without amyloid movement?", "why_it_matters": "Would separate a downstream synapse-protection axis from amyloid-burden claims.", "what_would_move_it": "A complement-directed Alzheimer's trial that pairs a synaptic-PET (SV2A) endpoint with amyloid imaging - none is currently registered." }, { "id": "GAP-004", "priority": "open", "question": "Does infectious or immune-history signal survive prospective mechanistic timing?", "why_it_matters": "Keeps the one canonical marginal lane alive without promoting association into causation.", "what_would_move_it": "Prospective cohort with reactivation markers, vaccination timing, negative controls, and dementia subtype resolution." }, { "id": "GAP-005", "priority": "provenance", "question": "Which public figures and rate claims have source-script-grade provenance?", "why_it_matters": "Prevents attractive SOTA summaries from laundering weak extraction chains.", "what_would_move_it": "Reproducible source scripts, assay metadata, cohort IDs, and four-number rate reports." }, { "id": "GAP-006", "priority": "instrument", "question": "Does the closure-role classifier agree with a genuinely blind grader at >= 95%?", "why_it_matters": "The public rate claims are only as trustworthy as the instrument; a blind July audit put agreement at 56%, so F_audit is TRIGGERED and no corrected rate ships.", "what_would_move_it": "A v0.6 precision revision aimed at amyloid over-firing, validated against the reserved 20% holdout - not tuned on the audited sample." } ], "predictions": [ { "id": "F1", "status": "open · no adequately-powered vehicle", "bet": "High-WMH tertile carries >60% of anti-amyloid tau-PET benefit.", "falsifier": "WMH stratification shows no enrichment, opposite direction, or only artifact after baseline severity adjustment. (Path note: the trials named for this test either record no tau or exclude high-WMH patients.)" }, { "id": "F2 ★★", "status": "open · decisive · assay not yet built", "bet": "pro-CatD/total inflects >= 5 years before Ab42/40 in DIAN.", "falsifier": "Inflection is within +/-3 EYO of amyloid change, absent, or later than MDH2/retromer-linked anchors. (Path note: zero published assays measure pro-cathepsin-D in Alzheimer's CSF; the assay is the bottleneck, not cohort access.)" }, { "id": "F3", "status": "open · read out without the subgroup", "bet": "GLP-1 helps non-diabetic normoglycemic AD if the metabolic axis is independently causal.", "falsifier": "Benefit collapses after glycemic, vascular, weight, or survival stratification. (Path note: EVOKE published 2026-03 with no pre-specified normoglycemic subgroup and no HbA1c collection.)" }, { "id": "F4", "status": "open · no registered AD path", "bet": "Complement inhibition preserves synapses without requiring amyloid change.", "falsifier": "Synapse benefit tracks only amyloid movement, nonspecific inflammation, or baseline severity. (Path note: the antibody named for this bet has no Alzheimer's trial; unfalsifiable as written until a vehicle exists.)" }, { "id": "F5", "status": "bench-runnable · not yet run", "bet": "Retromer restoration can rescue Ab in human-neuron contexts.", "falsifier": "SORL1/retromer rescue fails under knockout/rescue controls or cannot dissociate from generic toxicity. (Path note: a clinical trial cannot resolve this; the TPT-172-class chaperone is real but the isogenic-neuron experiment has not been run.)" }, { "id": "F6", "status": "open · may be mis-specified", "bet": "Sleep-amyloid coupling is production-dominant rather than only clearance-dominant.", "falsifier": "SILK kinetics show clearance-only behavior under controlled sleep perturbation. (Path note: the one human SILK dataset computed production by assuming clearance is unchanged, so the >=2x ratio is undefined when clearance does not move - the prediction likely needs reformulating.)" }, { "id": "F7", "status": "open · cohorts lack the ERF panel", "bet": "MAPT-tau propagation can proceed without an ERF precursor in non-AD tauopathy contexts.", "falsifier": "Longitudinal ERF markers consistently precede tau propagation across MAPT/FTLD cohorts. (Path note: ALLFTD is the right population but registers neither ERF markers nor tau-PET; GENFI is not in the trial registry.)" }, { "id": "F8", "status": "open · OLEs omit the needed conjunction", "bet": "Anti-amyloid slowing plateaus in the 23-45% range unless an upstream axis is also moved.", "falsifier": "A clean trial breaches the ceiling on propagation/cognition without hidden subgroup or geometric artifact. (Path note: the named open-label extensions read out but no single cohort holds >3yr CDR-SB slowing and tau-PET propagation velocity together.)" }, { "id": "F9", "status": "open · bench-runnable now", "bet": "Fasudil's Ab-lowering runs through the SORL1/retromer handle (ERF), not a parallel route.", "falsifier": "Fasudil lowers Ab within 25% of wild-type effect in isogenic SORL1-knockout neurons (SORL1-independent), or the clinical benefit tracks blood-pressure/perfusion rather than genotype. (Path note: the decisive isogenic-neuron experiment is runnable today and has zero prior literature; the named trial reads out ~2028 but collects no genotype.)" } ], "claims": [ { "id": "KRAM-001", "status": "active", "label": "Amyloid is real, but not the whole cause", "claim": "Kram still puts the amyloid cascade in class R5b: a real signal and a real amplifier, but not a complete cause. On the 47-fold larger paper set its escape-pressure eased from 0.53 to 0.43. The structure held; the exact number moved.", "falsifier": "A clean ceiling-breach trial or mechanism-timed dataset shows amyloid removal alone halts upstream propagation rather than only slowing downstream damage.", "next": "Track ceiling breaches, tau-PET slope, subgroup stratification, and upstream-axis movement together.", "weight": "escape-pressure 0.43" }, { "id": "KRAM-002", "status": "decisive gap", "label": "ERF temporal precedence", "claim": "Endolysosomal/retromer failure is a candidate convergence node that could precede canonical amyloid decline.", "falsifier": "No EYO lead in DIAN, or MDH2 / retromer-linked axes deviate earlier under matched cohort analysis.", "next": "Build a validated isoform-resolved pro-CatD CSF assay (none exists in Alzheimer's yet), then run it on banked DIAN CSF with SEC fractionation and mutation-stratified timing.", "weight": "F2 ★★" }, { "id": "KRAM-005", "status": "audit triggered", "label": "The instrument is under audit, not cleared", "claim": "Kram's first genuinely blind forensic audit, on the expanded corpus, put closure-role agreement with a cold grader at 56% - so F_audit is TRIGGERED. The earlier 96.7% figure was measured with the classifier's own answer visible to the grader and does not transfer. A v0.5 precision revision was benchmark-clean but improved blind agreement by only one paper on a fresh sample; amyloid over-firing is the dominant remaining error.", "falsifier": "A blind re-audit on held-out papers reaches >= 95% agreement under a precision-revised classifier.", "next": "v0.6 classifier revision targeting amyloid-as-context over-firing; re-audit against the reserved 20% holdout, not the sample it was tuned on.", "weight": "F_audit 56% blind" }, { "id": "KRAM-006", "status": "path audit", "label": "No starred prediction has a registered path to an answer", "claim": "A resolution-path audit of all 9 starred predictions against the trial registry (51 trial IDs checked, 0 fabricated) found that none currently has a registered study sufficient to resolve it. Named datasets are often the wrong instrument: the amyloid-x-vascular trial records no tau and excludes high-WMH patients; the complement antibody has no Alzheimer's program; the GLP-1 readout already published without the needed subgroup.", "falsifier": "A registered study is identified (or launched) that collects the exact measurement a starred prediction requires, in one adequately-powered cohort.", "next": "Prioritize the bench-runnable tests (F5, F9 isogenic-neuron experiments) and the missing assays (F2 pro-CatD CSF) over waiting on trials that cannot answer the question.", "weight": "0 / 9 registered paths" }, { "id": "KRAM-003", "status": "watching", "label": "Infectious etiology remains open", "claim": "Infectious/immune-history evidence is strong enough to watch and too weak to promote as causal closure.", "falsifier": "Prospective mechanistic cohorts erase the association after subtype, vascular, and healthcare-access controls.", "next": "Require mechanism-timed replication before promotion.", "weight": "A6 marginal" }, { "id": "KRAM-004", "status": "boundary", "label": "Personal stories can raise questions, not settle them", "claim": "Family fear and lived anecdotes may nominate mechanisms, but cannot count as mechanism evidence until operationalized.", "falsifier": "A claim cannot be translated into measurable cohort, assay, confound, or perturbation terms.", "next": "Route personal narratives into variables, controls, and safe research questions.", "weight": "Debbe's Lens" } ], "streams": [ { "lane": "new papers", "signal": "Classify fresh dementia papers into closure roles; flag empty cells, contradiction pressure, and falsifier-trigger hits. Corpus now 4,425 role-carrying papers (5,965 total), spanning 1990-2026." }, { "lane": "public claims", "signal": "Convert cure/prevention/biomarker claims into evidence grade, missing controls, and falsifier." }, { "lane": "decisive experiments", "signal": "Keep F1-F9 visible with their real resolution paths - including which ones no registered trial can currently answer - so effort goes where evidence can actually move." }, { "lane": "gap pressure", "signal": "Show under-tested lanes and missing instruments (assays, cohorts) where competent effort has higher marginal value than another review of the favorite scalar." }, { "lane": "Debbe's Lens", "signal": "Urgency is honored as a search constraint, never as permission to overclaim." } ] }